CN116881624A - Composite extreme event forecasting method, device, computer equipment and storage medium - Google Patents

Composite extreme event forecasting method, device, computer equipment and storage medium Download PDF

Info

Publication number
CN116881624A
CN116881624A CN202311141790.0A CN202311141790A CN116881624A CN 116881624 A CN116881624 A CN 116881624A CN 202311141790 A CN202311141790 A CN 202311141790A CN 116881624 A CN116881624 A CN 116881624A
Authority
CN
China
Prior art keywords
extreme event
target
target compound
forecasting
event
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202311141790.0A
Other languages
Chinese (zh)
Other versions
CN116881624B (en
Inventor
郝增超
张璇
郝芳华
付永硕
何佳
郝莹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Normal University
Original Assignee
Beijing Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Normal University filed Critical Beijing Normal University
Priority to CN202311141790.0A priority Critical patent/CN116881624B/en
Publication of CN116881624A publication Critical patent/CN116881624A/en
Application granted granted Critical
Publication of CN116881624B publication Critical patent/CN116881624B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Optimization (AREA)
  • Operations Research (AREA)
  • Strategic Management (AREA)
  • Pure & Applied Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Mathematical Analysis (AREA)
  • Economics (AREA)
  • Educational Administration (AREA)
  • Computational Mathematics (AREA)
  • Development Economics (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Algebra (AREA)
  • Game Theory and Decision Science (AREA)
  • Probability & Statistics with Applications (AREA)
  • Marketing (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application relates to a compound extreme event forecasting method, a compound extreme event forecasting device, computer equipment and a storage medium. The method comprises the following steps: selecting a forecasting factor of the target compound extreme event from potential influence factors based on the potential influence factors of the target compound extreme event and a target compound index corresponding to the target compound extreme event constructed in advance; respectively inputting the current values of the forecasting factors into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models; performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event; the target forecast results are used for representing the intensity of the target compound extreme event occurring in the forecast period. The method can be used for forecasting the compound extreme event.

Description

Composite extreme event forecasting method, device, computer equipment and storage medium
Technical Field
The application relates to the technical field of weather hydrologic forecasting, in particular to a compound extreme event forecasting method, a device, computer equipment, a storage medium and a computer program product.
Background
Extreme weather events and extreme climate events are collectively referred to as extreme events, which refer to an abnormal weather or climate event that has a low probability of occurrence. Composite extreme events refer to a combination of multiple drivers and/or disasters that lead to social or environmental risks, including composite events over time, i.e., occurring simultaneously or consecutively at the same site; and spatially compounded events, i.e., multiple sites occur simultaneously. In the context of global warming, extreme events are frequent and concurrent. In recent years, a plurality of areas are affected by compound extreme events such as high temperature drought, etc., so that the water supply safety, the grain production, the human health, etc. are seriously affected, and the ecological system is also deeply affected. Therefore, the method has great significance for analysis and prediction of compound extreme events.
Therefore, a method for forecasting the complex extreme event is needed.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a composite extreme event prediction method, apparatus, computer device, computer readable storage medium, and computer program product that are capable of predicting composite extreme events.
In a first aspect, the present application provides a method of composite extreme event prediction. The method comprises the following steps:
Selecting a predictor of the target compound extreme event from the potential influencing factors based on the potential influencing factors of the target compound extreme event and a target compound index which is pre-constructed and corresponds to the target compound extreme event;
respectively inputting the current value of the forecasting factor into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models;
performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period.
In one embodiment, the performing bayesian averaging on the plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event includes:
determining posterior probability of a forecasting result of the compound extreme event forecasting model based on the training data set of the target compound extreme event aiming at each compound extreme event forecasting model;
Taking the posterior probability of the forecasting result of the compound extreme event forecasting model as the weight corresponding to the compound extreme event forecasting model;
and weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model.
In one embodiment, the selecting the predictor of the target compound extreme event based on the potential influence factor of the target compound extreme event and the target compound index corresponding to the target compound extreme event constructed in advance includes:
aiming at each potential influence factor of a target compound type extreme event, carrying out partial correlation analysis on the potential influence factors and a target compound index corresponding to the target compound type extreme event constructed in advance to obtain partial correlation coefficients of the potential influence factors;
and if the partial correlation coefficient is larger than a preset partial correlation coefficient threshold value, taking the potential influence factor as a forecasting factor of the target compound extreme event.
In one embodiment, the method further comprises:
Determining a plurality of index factors corresponding to the target compound extreme event;
and constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function.
In one embodiment, the target compound extreme event includes a plurality of target single extreme events, and the determining a plurality of exponential factors corresponding to the target compound extreme event includes:
aiming at each target single extreme event, determining a target factor corresponding to the target single extreme event in preset influence factors;
and forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events.
In one embodiment, the constructing the target composite index corresponding to the target composite extreme event based on the plurality of index factors and the joint distribution function includes:
calculating a marginal distribution result of each index factor;
calculating a multi-element joint probability corresponding to the target compound extreme event based on marginal distribution results and joint distribution functions of the index factors;
and carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event.
In one embodiment, the method further comprises:
acquiring an evaluation data set; the evaluation data set comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event;
based on the evaluation data set, a leave-one-out cross-validation method is adopted to obtain a plurality of target forecast values of the target compound extreme event; based on the evaluation data set and a preset autoregressive model, a leave-one-out cross validation method is adopted to obtain a plurality of reference forecast values of the target compound extreme event;
calculating a first evaluation value based on a historical value of a target composite index corresponding to the target composite extreme event, a plurality of target forecast values of the target composite extreme event and a preset model evaluation index; calculating a second evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of reference forecast values of the target compound extreme event and a preset model evaluation index;
and if the first evaluation value and the second evaluation value do not meet a preset evaluation passing condition, updating the bias correlation coefficient threshold value, returning to a potential influence factor based on the target compound extreme event and a target compound index corresponding to the target compound extreme event which is constructed in advance, and selecting a forecasting factor of the target compound extreme event from the potential influence factors.
In a second aspect, the application also provides a compound extreme event forecasting device. The device comprises:
the selection module is used for selecting a forecasting factor of the target compound extreme event based on the potential influence factor of the target compound extreme event and a target compound index corresponding to the target compound extreme event, which is constructed in advance;
the first forecasting module is used for respectively inputting the current value of the forecasting factor into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models;
the second forecasting module is used for carrying out Bayesian averaging on a plurality of forecasting results of the target compound extreme event to obtain a target forecasting result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period.
In one embodiment, the second forecasting module is specifically configured to:
determining posterior probability of a forecasting result of the compound extreme event forecasting model based on the training data set of the target compound extreme event aiming at each compound extreme event forecasting model;
Taking the posterior probability of the forecasting result of the compound extreme event forecasting model as the weight corresponding to the compound extreme event forecasting model;
and weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model.
In one embodiment, the selecting module is specifically configured to:
aiming at each potential influence factor of a target compound type extreme event, carrying out partial correlation analysis on the potential influence factors and a target compound index corresponding to the target compound type extreme event constructed in advance to obtain partial correlation coefficients of the potential influence factors;
and if the partial correlation coefficient is larger than a preset partial correlation coefficient threshold value, taking the potential influence factor as a forecasting factor of the target compound extreme event.
In one embodiment, the apparatus further comprises:
the determining module is used for determining a plurality of index factors corresponding to the target compound extreme event;
and the construction module is used for constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function.
In one embodiment, the target compound extreme event includes a plurality of target single extreme events, and the determining module is specifically configured to:
aiming at each target single extreme event, determining a target factor corresponding to the target single extreme event in preset influence factors;
and forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events.
In one embodiment, the construction module is specifically configured to:
calculating a marginal distribution result of each index factor;
calculating a multi-element joint probability corresponding to the target compound extreme event based on marginal distribution results and joint distribution functions of the index factors;
and carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event.
In one embodiment, the apparatus further comprises:
the acquisition module is used for acquiring the evaluation data set; the evaluation data set comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event;
The verification module is used for obtaining a plurality of target forecast values of the target compound extreme event by adopting a leave-one-out cross verification method based on the evaluation data set; based on the evaluation data set and a preset autoregressive model, a leave-one-out cross validation method is adopted to obtain a plurality of reference forecast values of the target compound extreme event;
the calculation module is used for calculating a first evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of target forecast values of the target compound extreme event and a preset model evaluation index; calculating a second evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of reference forecast values of the target compound extreme event and a preset model evaluation index;
and the updating module is used for updating the bias correlation coefficient threshold value and returning the potential influence factor based on the target compound extreme event and the target compound index corresponding to the target compound extreme event which is constructed in advance if the first evaluation value and the second evaluation value do not meet the preset evaluation passing condition, and selecting the forecasting factor of the target compound extreme event from the potential influence factors.
In a third aspect, the present application also provides a computer device. The computer device comprises a memory storing a computer program and a processor implementing the steps of the first aspect described above when the processor executes the computer program.
In a fourth aspect, the present application also provides a computer-readable storage medium. The computer readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the first aspect described above.
In a fifth aspect, the present application also provides a computer program product. The computer program product comprising a computer program which, when executed by a processor, carries out the steps of the first aspect described above.
The composite extreme event forecasting method, the composite extreme event forecasting device, the computer equipment, the storage medium and the computer program product are characterized in that the forecasting factors of the target composite extreme event are selected based on potential influence factors of the target composite extreme event and target composite indexes corresponding to the target composite extreme event, which are constructed in advance, in the potential influence factors; respectively inputting the current value of the forecasting factor into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models; performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period. Thus, a target compound index corresponding to the target compound extreme event is pre-constructed, a forecasting factor is selected from potential influence factors corresponding to the target compound extreme event based on the target compound index, the current value of the forecasting factor is respectively input into a plurality of machine learning models which are trained, bayesian averaging is carried out on a plurality of forecasting results output by the plurality of machine learning models, a target forecasting result representing the occurrence intensity of the target compound extreme event in a forecasting period is obtained, and the compound extreme event is forecasted.
Drawings
FIG. 1 is a flow chart of a method for forecasting compound extreme events in one embodiment;
FIG. 2 is a flowchart illustrating a Bayesian averaging step performed on a plurality of forecast results for a target compound extreme event in one embodiment;
FIG. 3 is a flowchart illustrating a predictor step for selecting a target compound extreme event according to one embodiment;
FIG. 4 is a flow diagram of a process for constructing a target composite index in one embodiment;
FIG. 5 is a flow chart of determining a plurality of exponential factors corresponding to a target compound extreme event according to one embodiment;
FIG. 6 is a flowchart illustrating a step of constructing a target composite index corresponding to a target composite extreme event based on a plurality of index factors and a joint distribution function in one embodiment;
FIG. 7 is a flow chart of a method for forecasting compound extreme events according to another embodiment;
FIG. 8 is a block diagram of a composite extreme event forecasting device in one embodiment;
fig. 9 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
The present application will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present application more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the application.
In one embodiment, as shown in fig. 1, a compound extreme event prediction method is provided, where this embodiment is applied to a terminal to illustrate the method, and it is understood that the method may also be applied to a server, and may also be applied to a system including a terminal and a server, and implemented through interaction between the terminal and the server. The terminal can be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things equipment and portable wearable equipment, and the internet of things equipment can be smart speakers, smart televisions, smart air conditioners, smart vehicle-mounted equipment and the like. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The server may be implemented as a stand-alone server or as a server cluster composed of a plurality of servers. In this embodiment, the method includes the steps of:
step 101, selecting a forecasting factor of the target compound extreme event from potential influence factors based on the potential influence factors of the target compound extreme event and a target compound index corresponding to the target compound extreme event constructed in advance.
In the embodiment of the application, the target compound extreme event is a compound extreme event to be forecasted. Extreme weather events and extreme climate events are collectively referred to as extreme events, which refer to some abnormal weather or climate event that has a low probability of occurrence. Composite extreme events refer to a combination of multiple drivers and/or disasters that lead to social or environmental risks, including composite events over time, i.e., occurring simultaneously or consecutively at the same site; and spatially compounded events, i.e., multiple sites occur simultaneously. For example, extreme events may be high temperature events, drought events, and high wind events, and compound extreme events may be high temperature, drought, and high wind events that occur simultaneously, i.e., compound dry-hot wind events. The composite dry hot air can cause the influence of grain yield reduction and the like. The potential influencing factor is a factor capable of influencing the occurrence probability of the target compound type extreme event. Potential impact factors may include meteorological factors, terrestrial factors, and marine factors. For example, potential impact factors for a composite dry-hot wind event may include meteorological factors, land factors (also known as underlying data), which may include precipitation, air temperature, and wind speed, land factors, which may include soil moisture (or soil water), vegetation, and circulation indexes (also known as large scale circulation factors), which may include El Nino-southern billows (El Nino-Southern Oscillation, ENSO), pacific annual oscillations (Pacific Decadal Oscillation, PDO), and north atlantic billows (the North Atlantic Oscillation, NAO). The target compound index is an index corresponding to the target compound extreme event and is used for representing the intensity of the target compound extreme event occurring in the foreseeing period. The forecast period is a time period for forecasting the target compound extreme event, and can be a month scale, a week scale and a day scale. For example, the forecast period may be 1 month, 2 months, and 3 months into the future. For example, the target composite index corresponding to the composite dry-air event is a composite dry-air index (SDHWI) for characterizing the intensity of the composite dry-air event occurring during the forecast period. Thus, the composite high-temperature drought index can be used for evaluating agricultural production conditions and sand conditions of sand source lands. Under the background of climate warming, the method has a good indication effect on natural disasters such as sand storm, forest fire and the like in agricultural production management. The forecasting factor is a factor based on forecasting the target compound extreme event, and can be a potential influencing factor which has a larger influence on the occurrence of the target compound extreme event in the forecasting period.
The terminal builds a target composite index corresponding to the target composite extreme event in advance. And then, the terminal calculates the correlation degree between each potential influence factor and the target compound index based on the potential influence factor of the target compound extreme event and the target compound index corresponding to the pre-constructed target compound extreme event. The terminal then selects a predictor of the target composite extremity event among the potential impact factors based on the correlation between each potential impact factor and the target composite index.
Step 102, the current values of the forecasting factors are respectively input into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event.
The multiple composite extreme event forecasting models are realized by adopting different machine learning models.
In the embodiment of the application, aiming at each compound type extreme event forecasting model trained in advance, the terminal inputs the current value of the forecasting factor to the compound type extreme event forecasting model to obtain the forecasting result of the target compound type extreme event. Wherein the plurality of forecasted results for the target compound extreme event includes the forecasted results for each compound extreme event forecasting model. The forecast result includes a forecast value of the target composite index. The plurality of composite extreme event prediction models are implemented using a plurality of machine learning models. The multiple composite extreme event prediction models can be realized by adopting a linear regression model, a random forest model, a support vector machine number model, a deep learning model and the like. The current value of the predictor is the current grid point or site data of the predictor.
And step 103, performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event.
The target forecasting result is used for representing the intensity of the target compound extreme event occurring in the forecasting period.
In an embodiment of the application, the terminal trains a bayesian model average (BMA Bayesian model averaging, BMA) model of the target compound extreme event based on a training dataset of the target compound extreme event. And then, the terminal inputs a plurality of forecasting results of the target compound extreme event to a Bayesian model average model to obtain a target forecasting result of the target compound extreme event. Wherein the target forecast result comprises a target forecast value of the target composite index. The training data set is used for training a Bayesian model average model, and comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event.
In the composite extreme event forecasting method, a forecasting factor of the target composite extreme event is selected from potential influence factors based on the potential influence factors of the target composite extreme event and a target composite index corresponding to the target composite extreme event, which is constructed in advance; respectively inputting the current values of the forecasting factors into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models; performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event; the target forecast results are used for representing the intensity of the target compound extreme event occurring in the forecast period. Thus, a target compound index corresponding to the target compound extreme event is pre-constructed, a forecasting factor is selected from potential influence factors corresponding to the target compound extreme event based on the target compound index, the current value of the forecasting factor is respectively input into a plurality of machine learning models which are trained, bayesian averaging is carried out on a plurality of forecasting results output by the plurality of machine learning models, a target forecasting result representing the occurrence intensity of the target compound extreme event in a forecasting period is obtained, and the compound extreme event is forecasted. In addition, the method does not rely on a single machine learning model for forecasting, adopts a plurality of machine learning models for forecasting respectively, carries out Bayesian weighted average on each model, integrates a plurality of machine learning methods, integrates the advantages of the plurality of machine learning models, establishes an aggregate forecasting method, and can improve the accuracy of forecasting the compound extreme event. In addition, the method considers various forecasting signals from the early atmosphere, land and sea, has more comprehensive forecasting factors and can further improve the accuracy of forecasting the compound extreme event. In addition, the method belongs to a statistical method, is convenient to calculate, can effectively utilize multi-source data to forecast compound extreme events for a medium and long time, such as compound dry hot air events, and provides early warning information for disasters such as agricultural production and sand storm, so that effective preventive measures are taken to reduce related damages.
In one embodiment, as shown in fig. 2, the specific process of performing bayesian averaging on the multiple forecast results of the target compound extreme event to obtain the target forecast result of the target compound extreme event includes:
step 201, for each compound extreme event prediction model, determining a posterior probability of a prediction result of the compound extreme event prediction model based on a training data set of the target compound extreme event.
In the embodiment of the application, the terminal determines the posterior probability of the forecasting result of the composite extreme event forecasting model based on a Bayesian probability forecasting formula of the training data set of the target composite extreme event. Wherein the probability distribution parameters in the bayesian formulation can be estimated by a expectation maximization algorithm. The bayesian probability prediction formula can be expressed as:
wherein D is measured data (including the measured values of the predictor and the target complex index),a set of forecast results for n models; />Is the kth model forecast f k Posterior probability of (f) reflecting f k The degree of matching with the measured value of the target composite index, i.e. the weight w of the model k The model with better forecasting effect obtains higher weight; / >Is given f k And D, a forecast value of the target composite index under the condition.
Step 202, using posterior probability of the forecasting result of the composite extreme event forecasting model as the weight corresponding to the composite extreme event forecasting model.
In the embodiment of the application, the weight corresponding to the compound extreme event prediction model is used for representing the influence degree of the prediction result of the compound extreme event prediction model on the target prediction result of the target compound extreme event. It will be appreciated that for different foreseeable periods, the terminal may determine the weights corresponding to the respective composite extreme event prediction models for the different foreseeable periods, respectively, the determination process being similar to steps 201-202. The weights corresponding to the compound extreme event prediction models under different prediction periods can be the same or different.
And 203, weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model.
In the embodiment of the application, aiming at each compound extreme event prediction model, the terminal multiplies the prediction result of the compound extreme event prediction model by the weight corresponding to the compound extreme event prediction model to obtain the target product of the compound extreme event prediction model. And then, adding the target products of the composite extreme event prediction models by the terminal to obtain a target prediction result of the target composite extreme event.
In one embodiment, the terminal may calculate the target prediction result of the target composite extreme event through a posterior distribution mean formula of bayesian set prediction. The posterior distribution mean formula of bayesian set prediction can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,the posterior distribution mean value of Bayesian set prediction is also the target prediction result, w, of the target compound extreme event k Is the weight of the kth model, f k Is the forecast result of the kth model.
In the compound extreme event forecasting method, for each compound extreme event forecasting model, the posterior probability of the forecasting result of the compound extreme event forecasting model is determined based on the training data set of the target compound extreme event; the posterior probability of the forecasting result of the composite extreme event forecasting model is used as the weight corresponding to the composite extreme event forecasting model; and weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model. In this way, the prediction results of the multiple composite extreme event prediction models are subjected to Bayesian weighted average, so that the advantages of the multiple models are integrated under the condition that the prediction accuracy of each model is unknown and the model is not determined to be optimal, and the accuracy of the composite extreme event prediction can be improved.
In one embodiment, as shown in fig. 3, based on the potential impact factors of the target compound extreme event and the target compound index corresponding to the pre-constructed target compound extreme event, the specific process of selecting the predictor of the target compound extreme event among the potential impact factors includes the following steps:
step 301, performing partial correlation analysis on each potential influence factor of the target compound extreme event and a target compound index corresponding to the pre-constructed target compound extreme event to obtain a partial correlation coefficient of the potential influence factor.
In the embodiment of the application, when researching the influence or the correlation degree of a certain potential influence factor on a forecast object (namely, a target composite index), the influence of other potential influence factors is regarded as constant or unchanged, namely, the influence of other potential influence factors is not considered temporarily, only the degree of closeness of the correlation between the potential influence factor and the forecast object is researched, and the obtained numerical result is the partial correlation coefficient.
Aiming at each potential influence factor of the target compound type extreme event, the terminal performs partial correlation analysis on the potential influence factor and the target compound index corresponding to the target compound type extreme event based on the first historical value of the potential influence factor and the second historical value of the target compound index corresponding to the target compound type extreme event, which are built in advance, so as to obtain the partial correlation coefficient of the potential influence factor. The first history time corresponding to the first history value is earlier than the second history time corresponding to the second history value, and the first history time and the second history value differ by a preset foreseeing period.
Step 302, if the bias correlation coefficient is greater than the preset bias correlation coefficient threshold, using the potential influence factor as a predictor of the target compound extreme event.
It will be appreciated that for different foreseeable periods, the terminal may determine the predictors of the target compound extreme event under the different foreseeable periods, respectively, the process of determining being similar to steps 301-302. The predictors of the target compound extreme events under different foreseeable periods can be the same or different.
In the method for forecasting the composite extreme event, partial correlation analysis is carried out on each potential influence factor of the target composite extreme event and a target composite index corresponding to the target composite extreme event constructed in advance to obtain a partial correlation coefficient of the potential influence factor; and if the partial correlation coefficient is larger than a preset partial correlation coefficient threshold value, using the potential influence factor as a forecasting factor of the target compound extreme event. In this way, partial correlation analysis is carried out on each potential influence factor of the target composite index and the target composite extreme event, the degree of closeness of the correlation between the potential influence factors and the target composite index is determined, and only the potential influence factors which are high in the degree of closeness and obviously related to the target composite index are used as the forecasting factors for forecasting the composite extreme event, so that the target composite index is forecasted through the forecasting factors, and the accuracy of forecasting the composite extreme event can be improved.
In one embodiment, as shown in fig. 4, the compound extreme event prediction method further includes the following steps:
in step 401, a plurality of exponential factors corresponding to the target compound extreme event are determined.
In the embodiment of the application, the index factor is an influence factor of a target compound index corresponding to a target compound extreme event, is generated on the target compound extreme event, has a target influence degree, and can be set by a technician according to experience.
Step 402, constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function.
In the embodiment of the application, the terminal can calculate the multi-element joint probability of the target compound extreme event based on a plurality of exponential factors and joint distribution functions. Then, the terminal constructs a target composite index corresponding to the target composite extreme event based on the multi-element joint probability of the target composite extreme event. Wherein the joint distribution function may be a Copula function.
In the composite extreme event forecasting method, a plurality of index factors corresponding to the target composite extreme event are determined; and constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function. Therefore, based on the generation of the target compound extreme event, a plurality of index factors with target influence degree and the combined distribution function are used for constructing a target compound index, the target compound index is defined by combining the variable through the combined distribution function, and further, the compound extreme event forecast is realized, and the accuracy of the compound extreme event forecast can be improved by constructing the target compound index.
In one embodiment, as shown in fig. 5, the target compound extreme event includes a plurality of target single extreme events, and the specific process of determining a plurality of exponential factors corresponding to the target compound extreme event includes the following steps:
in step 501, for each target single extreme event, a target factor corresponding to the target single extreme event is determined in a preset influence factor.
In an embodiment of the present application, the target compound extreme event is composed of a plurality of target single extreme events. For example, a composite dry and hot wind event includes three targeted single extreme events of drought, high temperature and high winds. The target factor is an influence factor having a target influence degree on the generation of a target single extreme event, and can be set by a technician according to experience. For example, the target factor corresponding to the drought event is soil humidity, the target factor corresponding to the high temperature event is air temperature, and the target factor corresponding to the high wind event is wind speed. A target single extreme event corresponds to a target factor. Since the target compound extreme event is composed of a plurality of target single extreme events, one target compound extreme event corresponds to a plurality of exponential factors. The impact factor may be the total impact factor that affects the extreme event.
Step 502, forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events.
In one embodiment, the target factor corresponding to the drought event is soil humidity, the target factor corresponding to the high temperature event is air temperature, and the target factor corresponding to the high wind event is wind speed. And the terminal forms a plurality of index factors corresponding to the composite dry and hot air event by using the soil humidity, the air temperature and the air speed.
In the above-mentioned compound extreme event prediction method, for each target single extreme event, determining a target factor corresponding to the target single extreme event in a preset influence factor; and forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events. In this way, the target factors corresponding to the single extreme events included in the target compound extreme event are formed into a plurality of index factors corresponding to the target compound extreme event, and then the construction of the target compound index is realized.
In one embodiment, as shown in fig. 6, the specific process of constructing the target composite index corresponding to the target composite extreme event based on the multiple index factors and the joint distribution function includes the following steps:
Step 601, for each index factor, calculating a marginal distribution result of the index factor.
In the embodiment of the application, when the terminal calculates the marginal distribution result of each exponential factor, namely when the terminal fits the edge distribution to each variable, a non-parameter kernel density estimation method can be adopted for estimation. This eliminates the need to assume the distribution of variables.
Step 602, calculating a multi-element joint probability corresponding to the target compound extreme event based on the marginal distribution result and the joint distribution function of each exponential factor.
In the embodiment of the application, under the condition that the target compound type extreme event comprises three target single extreme events, the terminal calculates a two-dimensional joint distribution result and a three-dimensional joint distribution result corresponding to the target compound type extreme event based on the marginal distribution result and the joint distribution function of each exponential factor. And then, the terminal calculates the multi-element joint probability corresponding to the target compound extreme event based on the marginal distribution result of the exponential factor, the two-dimensional joint distribution result and the three-dimensional joint distribution result corresponding to the target compound extreme event. Wherein the multivariate joint probability is a non-standard variable. When the terminal calculates a two-dimensional joint distribution result corresponding to the target compound type extreme event, namely when two-dimensional joint distribution is simulated, a Copula function is selected from Gaussian, student-t, clayton, gumbel, frank and Joe Copula, and the Copula family is widely used in hydrological research. At each lattice or site, the parameters of the Copula function are estimated by maximum likelihood estimation and the most suitable Copula function is selected according to Akaike information criteria (Akaike information criterion, AIC). The three-dimensional joint distribution function is constructed by a Vine Copula function. Conventional multi-element Copula has some limitations, such as poor flexibility in modeling high-dimensional correlations, and the Vine Copula function can overcome the limitations and has been widely used in various fields. The three-dimensional joint probability is calculated here by determining the best fit two-dimensional Copula and its corresponding parameters to construct the C-Vine Copula.
In one embodiment, the terminal calculates a multivariate joint probability corresponding to the composite dry-hot wind event, which can be expressed as:
wherein X, Y, Z represents soil humidity, air temperature and wind speed, respectively; u=f X (x)、v=F Y (y)、w=F Z (Z) is the marginal distribution result of X, Y and Z, respectively; c is a Copula function; c (u, v) and C (u, w) are two-dimensional joint distribution results corresponding to the composite dry and hot air event, and C (u, v, w) is a three-dimensional joint distribution result corresponding to the composite dry and hot air event.
And 603, carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event.
In the embodiment of the application, the terminal performs standardized conversion on the multi-element joint probability, converts a non-standard variable into a standard variable, and obtains a target composite index corresponding to a target composite extreme event. Wherein the target composite index is a standard variable.
In one embodiment, the terminal performs standardization processing on the multiple joint probabilities to obtain a composite dry-hot air index corresponding to the composite dry-hot air event, which can be expressed as:
wherein P is a multivariate joint probability; n (N) -1 Is an inverse function of the standard normal distribution. The lower the composite dry hot air index value, the more severe the composite dry hot air event.
In the above-mentioned compound extreme event prediction method, for each index factor, calculating the marginal distribution result of the index factor; calculating a multi-element joint probability corresponding to the target compound extreme event based on marginal distribution results and joint distribution functions of the index factors; and (3) carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event. Therefore, based on the joint distribution function, the multi-element joint probability corresponding to the target compound extreme event is calculated, and the multi-element joint probability is subjected to standardized conversion, so that the occurrence intensity of the target compound extreme event in the prediction period can be effectively represented, and the accuracy of the compound extreme event prediction is improved.
In one embodiment, as shown in fig. 7, the compound extreme event prediction method further includes the following steps:
step 701, an evaluation dataset is acquired.
The evaluation data set comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event.
In the embodiment of the application, the evaluation data set may be the same as or different from the training data set of the target compound extreme event.
Step 702, obtaining a plurality of target forecast values of the target compound extreme event by adopting a leave-one-out cross-validation method based on the evaluation data set. Based on the evaluation data set and a preset autoregressive model, a leave-one-out cross-validation method is adopted to obtain a plurality of reference forecast values of the target compound extreme event.
In the embodiment of the application, the terminal divides the data of the evaluation data set into N groups, randomly selects one group from the N groups of data as a test set, and takes the rest N-1 groups as a training set. The terminal then trains each composite extreme event prediction model based on the training set data. And then, the terminal respectively inputs the historical values of the forecasting factors of the target compound extreme events included in the test set into each compound extreme event forecasting model after the training is finished, and a forecasting value corresponding to each compound extreme event forecasting model is obtained. Then, the terminal selects the next set of data of the test set as a new test set, and returns to the step of taking the remaining N-1 set as a training set. And the terminal obtains N target forecast values of the target compound extreme event and N forecast values corresponding to the compound extreme event forecast models until the cycle times are N times. Since the sample size of the season scale forecast is relatively small, the data can be fully utilized. Then, the terminal adopts the compound extreme event forecasting method to carry out Bayesian average on the forecasting results of a plurality of groups of target compound extreme events, so as to obtain a plurality of target forecasting values of the target compound extreme events. Meanwhile, the terminal obtains a plurality of reference forecast values of the target compound extreme event by adopting a leave-one-out cross validation method based on the evaluation data set and a preset autoregressive model. The autoregressive model is built based on the self continuity of the composite index sequence corresponding to the training set.
In step 703, a first evaluation value is calculated based on the historical value of the target composite index corresponding to the target composite extreme event, the plurality of target forecast values of the target composite extreme event, and the preset model evaluation index. And calculating a second evaluation value based on the historical value of the target compound index corresponding to the target compound extreme event, a plurality of reference forecast values of the target compound extreme event and a preset model evaluation index.
In the embodiment of the application, the model evaluation index is an index for evaluating the accuracy of the model and is used for evaluating the degree of difference between the predicted value and the observed value (true value). Model evaluation indexes can be Nash efficiency coefficient (Nash-Sutcliffe efficiency coefficient, NSE) and Root Mean square error (Root Mean SquaredError, RMSE). The first evaluation value and the second evaluation value are both values of the model evaluation index. The Nash efficiency coefficient and the root mean square error calculation formula are respectively:
wherein i is a forecasting period, i.e. an i-th group of test sets, and n is a forecasting total period number, i.e. a total group number of the test sets; s is S p And S is o Is the forecast value and the observed value of the target composite index;is the mean of the observations. NSE ranges from- ≡1. Nse=1 represents a perfect forecast. NSE near 0 indicates that the prediction result is similar to the mean value of the observed values (but with larger error), NSE <0 represents an untrusted forecasting result. RMSE ranges from 0 to +.infinity, rmse=0 indicates perfect prediction. Higher NSE and lower RMSE indicate better forecasting performance.
And step 704, if the first evaluation value and the second evaluation value do not meet the preset evaluation passing condition, updating the bias correlation coefficient threshold value, returning to a target compound index corresponding to the target compound extreme event based on the potential influence factor of the target compound extreme event and the pre-constructed target compound extreme event, and selecting a forecasting factor of the target compound extreme event from the potential influence factors.
In an embodiment of the application, the terminal compares the second evaluation value with the first evaluation value. And under the condition that the model evaluation index is a Nash efficiency coefficient, if the first evaluation value is smaller than the second evaluation value, the terminal determines that the first evaluation value and the second evaluation value do not meet a preset evaluation passing condition. And under the condition that the model evaluation index is root mean square error, if the first evaluation value is larger than the second evaluation value, the terminal determines that the first evaluation value and the second evaluation value do not meet the preset evaluation passing condition. The evaluation passing condition can be a condition for measuring whether the composite extreme event forecasting method adopting the current forecasting factor is better than a preset autoregressive model. The evaluation passing condition corresponds to a model evaluation index.
In one example, if the first evaluation value and the second evaluation value do not satisfy the preset evaluation passing condition, the terminal lowers the bias correlation coefficient threshold. For example, the terminal subtracts a preset threshold change amount from the partial correlation coefficient threshold to obtain an updated partial correlation coefficient threshold. For example, the terminal multiplies the partial correlation coefficient threshold by a preset threshold change proportion to obtain an updated partial correlation coefficient threshold.
In the composite extreme event forecasting method, based on the same evaluation data set, the evaluation value of the preset model evaluation index is calculated to evaluate the set evaluation model and the preset autoregressive model of the method respectively, the evaluation can be performed in real time, periodically or when the evaluation condition is triggered, and the bias correlation coefficient threshold value is updated in time under the condition that the evaluation result of the set evaluation model of the method is worse than the evaluation result of the autoregressive model, so that the forecasting factor is updated in time, the optimization of the set evaluation model of the method is realized, the accuracy of composite extreme event forecasting can be further improved, and the influence on the composite extreme event forecasting caused by different forecasting periods, different areas or other factors can be dealt with. In addition, the model is evaluated by adopting a leave-one-out cross-validation method, so that the data can be fully utilized, the actual situation that the sample size of the seasonal scale forecast is relatively small can be more met, and the evaluation accuracy of the composite extreme event forecast method can be improved.
It should be understood that, although the steps in the flowcharts related to the embodiments described above are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps in the flowcharts described in the above embodiments may include a plurality of steps or a plurality of stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of the steps or stages is not necessarily performed sequentially, but may be performed alternately or alternately with at least some of the other steps or stages.
Based on the same inventive concept, the embodiment of the application also provides a compound extreme event forecasting device for realizing the compound extreme event forecasting method. The implementation of the solution provided by the device is similar to that described in the above method, so the specific limitation in the embodiment of the compound type extreme event prediction device or devices provided below may be referred to the limitation of the compound type extreme event prediction method hereinabove, and will not be repeated herein.
In one embodiment, as shown in fig. 8, there is provided a composite extreme event forecasting apparatus 800 comprising: a selection module 810, a first forecasting module 820, and a second forecasting module 830, wherein:
the selecting module 810 is configured to select a predictor of the target compound extreme event based on a potential impact factor of the target compound extreme event and a target compound index corresponding to the target compound extreme event, which is pre-constructed, in the potential impact factor;
a first forecasting module 820, configured to input current values of the predictors to a plurality of pre-trained composite extreme event forecasting models, respectively, to obtain a plurality of forecasting results of the target composite extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models;
the second prediction module 830 is configured to perform bayesian averaging on a plurality of prediction results of the target compound extreme event to obtain a target prediction result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period.
Optionally, the second forecasting module 830 is specifically configured to:
Determining posterior probability of a forecasting result of the compound extreme event forecasting model based on the training data set of the target compound extreme event aiming at each compound extreme event forecasting model;
taking the posterior probability of the forecasting result of the compound extreme event forecasting model as the weight corresponding to the compound extreme event forecasting model;
and weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model.
Optionally, the selecting module 810 is specifically configured to:
aiming at each potential influence factor of a target compound type extreme event, carrying out partial correlation analysis on the potential influence factors and a target compound index corresponding to the target compound type extreme event constructed in advance to obtain partial correlation coefficients of the potential influence factors;
and if the partial correlation coefficient is larger than a preset partial correlation coefficient threshold value, taking the potential influence factor as a forecasting factor of the target compound extreme event.
Optionally, the apparatus 800 further includes:
the determining module is used for determining a plurality of index factors corresponding to the target compound extreme event;
And the construction module is used for constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function.
Optionally, the target compound extreme event includes a plurality of target single extreme events, and the determining module is specifically configured to:
aiming at each target single extreme event, determining a target factor corresponding to the target single extreme event in preset influence factors;
and forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events.
Optionally, the construction module is specifically configured to:
calculating a marginal distribution result of each index factor;
calculating a multi-element joint probability corresponding to the target compound extreme event based on marginal distribution results and joint distribution functions of the index factors;
and carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event.
Optionally, the apparatus 800 further includes:
the acquisition module is used for acquiring the evaluation data set; the evaluation data set comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event;
The verification module is used for obtaining a plurality of target forecast values of the target compound extreme event by adopting a leave-one-out cross verification method based on the evaluation data set; based on the evaluation data set and a preset autoregressive model, a leave-one-out cross validation method is adopted to obtain a plurality of reference forecast values of the target compound extreme event;
the calculation module is used for calculating a first evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of target forecast values of the target compound extreme event and a preset model evaluation index; calculating a second evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of reference forecast values of the target compound extreme event and a preset model evaluation index;
and the updating module is used for updating the bias correlation coefficient threshold value and returning the potential influence factor based on the target compound extreme event and the target compound index corresponding to the target compound extreme event which is constructed in advance if the first evaluation value and the second evaluation value do not meet the preset evaluation passing condition, and selecting the forecasting factor of the target compound extreme event from the potential influence factors.
The above-mentioned various modules in the composite extreme event forecasting device may be implemented in whole or in part by software, hardware, and combinations thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure thereof may be as shown in fig. 9. The computer device includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input means. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input/output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program when executed by a processor implements a method of compound extreme event prediction. The display unit of the computer device is used for forming a visual picture, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be a key, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by persons skilled in the art that the architecture shown in fig. 9 is merely a block diagram of some of the architecture relevant to the present inventive arrangements and is not limiting as to the computer device to which the present inventive arrangements are applicable, and that a particular computer device may include more or fewer components than shown, or may combine some of the components, or have a different arrangement of components.
In one embodiment, a computer device is provided, comprising a memory and a processor, the memory having stored therein a computer program, the processor implementing the steps of the method embodiments described above when the computer program is executed.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the method embodiments described above.
In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the method embodiments described above.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and the data (including but not limited to data for analysis, stored data, presented data, etc.) related to the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the related laws and regulations and standards of the related country and region.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high density embedded nonvolatile Memory, resistive random access Memory (ReRAM), magnetic random access Memory (Magnetoresistive Random Access Memory, MRAM), ferroelectric Memory (Ferroelectric Random Access Memory, FRAM), phase change Memory (Phase Change Memory, PCM), graphene Memory, and the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory, and the like. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like. The databases referred to in the embodiments provided herein may include at least one of a relational database and a non-relational database. The non-relational database may include, but is not limited to, a blockchain-based distributed database, and the like. The processor referred to in the embodiments provided in the present application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, or the like, but is not limited thereto.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The foregoing examples illustrate only a few embodiments of the application and are described in detail herein without thereby limiting the scope of the application. It should be noted that it will be apparent to those skilled in the art that several variations and modifications can be made without departing from the spirit of the application, which are all within the scope of the application. Accordingly, the scope of the application should be assessed as that of the appended claims.

Claims (10)

1. A method of composite extreme event prediction, the method comprising:
selecting a predictor of the target compound extreme event from the potential influencing factors based on the potential influencing factors of the target compound extreme event and a target compound index which is pre-constructed and corresponds to the target compound extreme event;
respectively inputting the current value of the forecasting factor into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models;
Performing Bayesian averaging on a plurality of forecast results of the target compound extreme event to obtain a target forecast result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period.
2. The method of claim 1, wherein bayesian averaging the plurality of forecasted results for the target compound extremity event to obtain a target forecasted result for the target compound extremity event comprises:
determining posterior probability of a forecasting result of the compound extreme event forecasting model based on the training data set of the target compound extreme event aiming at each compound extreme event forecasting model;
taking the posterior probability of the forecasting result of the compound extreme event forecasting model as the weight corresponding to the compound extreme event forecasting model;
and weighting and calculating the target forecasting result of the target compound extreme event according to the forecasting result of each compound extreme event forecasting model and the weight corresponding to each compound extreme event forecasting model.
3. The method of claim 1, wherein the selecting the predictor of the target compound extremity event based on the potential impact factors of the target compound extremity event and a pre-constructed target compound index corresponding to the target compound extremity event comprises:
Aiming at each potential influence factor of a target compound type extreme event, carrying out partial correlation analysis on the potential influence factors and a target compound index corresponding to the target compound type extreme event constructed in advance to obtain partial correlation coefficients of the potential influence factors;
and if the partial correlation coefficient is larger than a preset partial correlation coefficient threshold value, taking the potential influence factor as a forecasting factor of the target compound extreme event.
4. The method according to claim 1, wherein the method further comprises:
determining a plurality of index factors corresponding to the target compound extreme event;
and constructing a target composite index corresponding to the target composite extreme event based on the index factors and the joint distribution function.
5. The method of claim 4, wherein the target compound extremity event comprises a plurality of target single extremity events, and wherein determining a plurality of exponential factors corresponding to the target compound extremity event comprises:
aiming at each target single extreme event, determining a target influence factor corresponding to the target single extreme event in preset influence factors;
and forming a plurality of index factors corresponding to the target compound extreme events by using the target factors corresponding to the target single extreme events.
6. The method of claim 4, wherein constructing a target composite index corresponding to the target composite extreme event based on the plurality of index factors and a joint distribution function comprises:
calculating a marginal distribution result of each index factor;
calculating a multi-element joint probability corresponding to the target compound extreme event based on marginal distribution results and joint distribution functions of the index factors;
and carrying out standardization processing on the multi-element joint probability to obtain a target composite index corresponding to the target composite extreme event.
7. A method according to claim 3, characterized in that the method further comprises:
acquiring an evaluation data set; the evaluation data set comprises a historical value of a predictor of the target compound extreme event and a historical value of a target compound index corresponding to the target compound extreme event;
based on the evaluation data set, a leave-one-out cross-validation method is adopted to obtain a plurality of target forecast values of the target compound extreme event; based on the evaluation data set and a preset autoregressive model, a leave-one-out cross validation method is adopted to obtain a plurality of reference forecast values of the target compound extreme event;
Calculating a first evaluation value based on a historical value of a target composite index corresponding to the target composite extreme event, a plurality of target forecast values of the target composite extreme event and a preset model evaluation index; calculating a second evaluation value based on a historical value of a target compound index corresponding to the target compound extreme event, a plurality of reference forecast values of the target compound extreme event and a preset model evaluation index;
and if the first evaluation value and the second evaluation value do not meet a preset evaluation passing condition, updating the bias correlation coefficient threshold value, returning to a potential influence factor based on the target compound extreme event and a target compound index corresponding to the target compound extreme event which is constructed in advance, and selecting a forecasting factor of the target compound extreme event from the potential influence factors.
8. A composite extreme event forecasting device, the device comprising:
the selection module is used for selecting a forecasting factor of the target compound extreme event based on the potential influence factor of the target compound extreme event and a target compound index corresponding to the target compound extreme event, which is constructed in advance;
The first forecasting module is used for respectively inputting the current value of the forecasting factor into a plurality of pre-trained compound extreme event forecasting models to obtain a plurality of forecasting results of the target compound extreme event; the multiple composite extreme event forecasting models are realized by adopting different machine learning models;
the second forecasting module is used for carrying out Bayesian averaging on a plurality of forecasting results of the target compound extreme event to obtain a target forecasting result of the target compound extreme event; the target forecast result is used for representing the intensity of the target compound extreme event occurring in the forecast period.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
CN202311141790.0A 2023-09-06 2023-09-06 Composite extreme event forecasting method, device, computer equipment and storage medium Active CN116881624B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311141790.0A CN116881624B (en) 2023-09-06 2023-09-06 Composite extreme event forecasting method, device, computer equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311141790.0A CN116881624B (en) 2023-09-06 2023-09-06 Composite extreme event forecasting method, device, computer equipment and storage medium

Publications (2)

Publication Number Publication Date
CN116881624A true CN116881624A (en) 2023-10-13
CN116881624B CN116881624B (en) 2023-11-17

Family

ID=88271945

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311141790.0A Active CN116881624B (en) 2023-09-06 2023-09-06 Composite extreme event forecasting method, device, computer equipment and storage medium

Country Status (1)

Country Link
CN (1) CN116881624B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117726038A (en) * 2023-12-26 2024-03-19 北京师范大学 Method and device for monitoring space composite high-temperature drought event and computer equipment

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113537600A (en) * 2021-07-20 2021-10-22 浙江省水利水电勘测设计院 Medium-and-long-term rainfall forecast modeling method based on whole-process coupled machine learning
CN113592144A (en) * 2021-06-28 2021-11-02 清华大学 Medium-and-long-term runoff probability forecasting method and system
CN113627690A (en) * 2021-09-03 2021-11-09 中国人民解放军国防科技大学 Method for predicting seasonal precipitation in southern China
CN114280696A (en) * 2021-12-23 2022-04-05 无锡九方科技有限公司 Intelligent sea fog level forecasting method and system
US11315046B1 (en) * 2018-06-28 2022-04-26 Ashton Robinson Cook Machine learning-based disaster modeling and high-impact weather event forecasting
CN115759445A (en) * 2022-11-24 2023-03-07 浙江大学 Machine learning and cloud model-based classified flood random forecasting method

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11315046B1 (en) * 2018-06-28 2022-04-26 Ashton Robinson Cook Machine learning-based disaster modeling and high-impact weather event forecasting
CN113592144A (en) * 2021-06-28 2021-11-02 清华大学 Medium-and-long-term runoff probability forecasting method and system
CN113537600A (en) * 2021-07-20 2021-10-22 浙江省水利水电勘测设计院 Medium-and-long-term rainfall forecast modeling method based on whole-process coupled machine learning
CN113627690A (en) * 2021-09-03 2021-11-09 中国人民解放军国防科技大学 Method for predicting seasonal precipitation in southern China
CN114280696A (en) * 2021-12-23 2022-04-05 无锡九方科技有限公司 Intelligent sea fog level forecasting method and system
CN115759445A (en) * 2022-11-24 2023-03-07 浙江大学 Machine learning and cloud model-based classified flood random forecasting method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZENGCHAO HAO等: "compound events and associated impacts in China", 《ISCIENCE》, vol. 25, no. 8, pages 1 - 10 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117726038A (en) * 2023-12-26 2024-03-19 北京师范大学 Method and device for monitoring space composite high-temperature drought event and computer equipment

Also Published As

Publication number Publication date
CN116881624B (en) 2023-11-17

Similar Documents

Publication Publication Date Title
Clark et al. The Schaake shuffle: A method for reconstructing space–time variability in forecasted precipitation and temperature fields
Robertson et al. A Bayesian approach to predictor selection for seasonal streamflow forecasting
Wu et al. Agricultural drought prediction based on conditional distributions of vine copulas
Van Huijgevoort et al. Global multimodel analysis of drought in runoff for the second half of the twentieth century
Goly et al. Development and evaluation of statistical downscaling models for monthly precipitation
Della-Marta et al. Improved estimates of the European winter windstorm climate and the risk of reinsurance loss using climate model data
Luomaranta et al. Multimodel estimates of the changes in the Baltic Sea ice cover during the present century
Alamgir et al. Downscaling and projection of spatiotemporal changes in temperature of Bangladesh
CN116881624B (en) Composite extreme event forecasting method, device, computer equipment and storage medium
CN114493052B (en) Multi-model fusion self-adaptive new energy power prediction method and system
Mohanty et al. Evaluation of performance of seasonal precipitation prediction at regional scale over India
Su et al. Multi‐site bias correction of climate model outputs for hydro‐meteorological impact studies: An application over a watershed in China
CN112585505A (en) Determining location-specific weather information for agronomic decision support
Keller et al. Downscaling approaches of climate change projections for watershed modeling: Review of theoretical and practical considerations
CN117909888B (en) Intelligent artificial intelligence climate prediction method
Gond et al. Assessment of drought variability using SPEI under observed and projected climate scenarios over Uttar Pradesh, India
Li et al. Improvements in subseasonal forecasts of rainfall extremes by statistical postprocessing methods
KR102496876B1 (en) Apparatus and method for forecasting the seasonal forest fire risk
CN116186493B (en) Meteorological forecasting method, device, equipment and medium based on numerical mode correction
Ren et al. Downscaling of daily extreme temperatures in the Yarlung Zangbo River Basin using machine learning techniques
CN110520887A (en) System and method for using the statistical analysis prediction economic trend of weather data
CN116796649A (en) SPEI coarse resolution data space downscaling method and device based on machine learning
CN116341841A (en) Runoff forecast error correction method, apparatus, device, medium and program product
Kaczmarska et al. Local generalised method of moments: an application to point process‐based rainfall models
Ghaemi et al. Uncertainty analysis of radar rainfall estimates over two different climates in Iran

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant